Chapter Two

Bernhard Hommel email address: hommel@fsw.leidenuniv.nl     Cognitive Psychology Unit & Leiden Institute for Brain and Cognition, Leiden University, Leiden, The Netherlands

Traditional approaches to action control assume the existence of a more or less unitary control system that struggles with, and serves to overcome action tendencies induced by automatic processes. In this article, I point out that and why these approaches fail to capture the complexity and dynamics of cognitive control. I describe an alternative approach that assumes that control emerges from the interaction of at least two counteracting forces: one system promoting persistence and the maintenance of action goals and another promoting mental and behavioral flexibility. I describe how this interaction might be shaped by various factors, including genetic predisposition, learning, personal experience, and the cultural context, and suggest a simple functional model (the Metacontrol State Model, MSM) that explains how this shaping process works. Then I provide an overview of studies from various fields (including perception, attention, performance monitoring, conflict resolution, creativity, meditation, religion, and interpersonal perception and behavior) that successfully tested predictions from the MSM.

Human action is often goal-oriented (to a degree that the existence of goals is often considered to be the defining aspect of human action), often initiated in the absence of action-related external stimulation, often spontaneous and nonhabitual, and can commonly be adjusted rapidly if no longer functional or not meeting situational challenges. The human capacity to achieve all of this is commonly ascribed to what we now call cognitive control or executive functions. The general idea is that the processes captured by these labels constitute a kind of second layer of information processing: while the first layer consists of basic processes to translate stimulation or input (the terms preferred by behavioristic and information-processing approaches, respectively)—such as sensory registration, identification, attentional selection, response decision, and motor execution—the second layer operates on the first by rewiring the respective processes in such a way that behavior is optimized and intended goals are reached (e.g., Verbruggen, McLaren, & Chambers, 2014).

Even though the terms cognitive control and executive function are relatively new (with Atkinson & Shiffrin, 1968; as one of the first proponents), the underlying concepts and the investigation of the underlying mechanisms are much older. In the following, I will discuss two predecessors and the related theoretical implications that have survived the terminological transition and that still dominate the discussion of human action control. Concepts that are used for such a long time and that have stimulated so much research are unlikely to be entirely incorrect, but I do think that they have limitations that need to be overcome in future research. I will thus present an alternative view of human action control, at least of some relevant aspects of it, that is also not without predecessors but has the advantage of providing a more dynamic, and thus probably more realistic, view of how humans keep their actions adaptive. I will begin by discussing two more traditional and popular views of action control—one assuming a will that is fighting against habits and another assuming an ego that mediates between societal requirements and personal urges—and point out a number of limitations of these views. Then I will synthesize the positive aspects of these two views and bring them together with more recent ideas from cognitive psychology and the cognitive neurosciences, to construct a more dynamic approach to human action control. Finally, I will sketch major determinants of the operation characteristics of control functions and provide some empirical examples to illustrate and justify my choices. However, I will not start without emphasizing that this chapter is a (empirical and theoretical) work in progress, a developing theoretical approach rather than a complete theory.

Approaches to human action control are older than the first psychological laboratory, which explains why they were commonly focusing on the conscious experience of control—the major empirical tool of nonexperimental researchers. The dominant view during the years in which the first laboratories were established was ideomotor theory, which has an even longer philosophical tradition (Stock & Stock, 2004). The approach focuses on the process of gaining control over one's body and its movements. It basically assumes that perceptual impressions of actively achieved changes in one's environment are automatically associated with whatever motor pattern was used to achieve them (for a review, see Hommel, 2009). Once these associations are created, the agent can reactivate the codes underlying the perceptual impressions at will and thereby automatically prime the associated motor patterns—the motor system thus becomes a slave of the will, as it were.

As the first experimental investigations of human cognition were focusing on perception, ideomotor theory and the will in general did not receive substantial empirical attention until the early 1900s. Around that time, experimental methods were developed to assess the time it takes to implement a particular action goal (e.g., Michotte & Prüm, 1911) and the impact of implementing a particular goal on information processing and conscious experience (Ach, 1910, 1935). Of particular interest, even the very first attempts to assess human will were using a scenario that survives to this day: Ach (1910) argued that the best and most objective way to measure the human will requires putting it in opposition to a counteracting force, which according to Ach was habit. Following this rationale, he developed methods to strengthen particular stimulus–response associations through repetition (e.g., by having participants produce rhyming responses to sequences of nonsense syllables) and then tested the degree of willpower by requiring the participant to produce a different response on old versus new items (for an overview, see Hommel, 2000a). As expected, items that were previously associated with a different action category were more difficult and took more time to respond to than new or familiar, but previously unassociated items. In numerous studies, Ach and colleagues demonstrated that the size of this interference effect varies systematically with the strength of the previously induced habit and/or the strength of the exerted or individually available willpower (for an overview, see Ach, 1935).

Note the similarity between this approach and the now popular rationale to assess the existence and efficiency of cognitive control functions by analyzing how much they are hampered by stimuli that are (or have been) associated with goal-incongruent responses—such as in Stroop, Simon, or flanker tasks (e.g., Hommel, 2000b). In fact, it is not only the experimental approach that is still heavily used, but the theoretical framework has remained in place as well. It comes in several flavors. The revival at the end of the 1960s along the lines of Atkinson and Shiffrin (1968) was basically replacing the apparently outdated terms “will” and “habit” by the more depersonalized terms “cognitive control” (or executive control, or executive function) and “automatic processes.” As in the original framework, novel and unfamiliar actions were supposed to rely on control processes to configure the cognitive system while overlearned actions could rely on automatic processes and thus require little or no control. The basic idea was that learning is the mechanism that creates automaticity over time by building more or less direct associations between stimulus representations and response representations (e.g., Kornblum, Hasbroucq, & Osman, 1990; de Wit & Dickinson, 2009). There is a long-standing controversy with regard to the question how automatic processes or automaticity need to be characterized, and it is fair to say that no universal criterion has been found so far. Some authors have advocated nonintentionality as the key criterion while others have suggested the ballistic, nonsuppressible nature of automatic processes, and even others have favored the lack of (commonly not further defined) “cognitive” or “central” resources needed to run automatic processes or the immunity to interference from other processes (e.g., Bargh, 1994; Moors & De Houwer, 2006).

The major problem underlying discussions and theoretical considerations of that sort is that they are facing a classical Rylean category issue (Ryle, 1949). This problem is a bit more transparent when one refers to the original concept of the “will.” Will is something that persons have but not a characteristic of a nervous system—much like there is nothing in a particular building that reflects its being part of a university, as in Ryle's famous example. Obviously, the willing person has a nervous system and (if we accept the functionalist stance underlying the rationale of cognitive psychology and the cognitive neurosciences) the process of willing...